Training the next generation in uniportal video-assisted thoracic surgery: A single-centre retrospective cohort study of operative and oncological outcomes.

BACKGROUND: Uniportal video-assisted thoracic surgery presents distinct technical demands that challenge traditional assumptions about surgical training. Evidence remains limited on how trainees are introduced to the technique in clinical practice and what outcomes follow. METHODS: A retrospective cohort study at Royal Papworth Hospital, including all uniportal VATS anatomical lung resections and thymectomies performed between May 2023 and August 2024; thymectomies were analysed separately. Cases were categorised as training, where the trainee was recorded as primary operator under consultant supervision, or non-training. The two groups differ as consultants assigned cases at their discretion according to perceived complexity. Mortality was followed to 8 August 2026. Post-operative morbidity was classified using the Clavien-Dindo system. Continuous differences are reported as Hodges-Lehmann location shifts with bootstrap confidence intervals. Operative time was modelled on the log scale, with confidence intervals bootstrapped over operators, and all comparisons were repeated excluding the highest-volume registrar. RESULTS: Out of 203 anatomical lung resections in 202 patients, 100 were training and 103 non-training. Measured baseline characteristics were similar, although the two groups differed in operative complexity. Operative time was longer in the training group but did not reach statistical significance (training median 2.32 versus 2.08 h; Hodges-Lehmann shift +0.20 h, 95% CI -0.08 to +0.52; p = 0.054). Adjusted for procedure type and supervising consultant, training cases were 7.4% longer (95% CI -1.0 to +16.6; p = 0.087); with confidence intervals bootstrapped over operators this widened to -3.6% to +22.5%. Post-operative length of stay was identical, with a median of 3 days in both groups (p = 0.872). There was one death within 30 days, in the non-training group, and one further death within 90 days, in the training group; these numbers are too small to support any comparison. Clavien-Dindo grade IIIa or above occurred in 10% of training versus 7.8% of non-training cases (risk difference +2.2%, 95% CI -5.6 to +10.1; p = 0.628), and prolonged air leak in 16% versus 10.7% (risk difference +5.3%, 95% CI -4.0 to +14.7; p = 0.305). Complete resection was achieved in 94.3% versus 90.4% (risk difference +3.9%, 95% CI -3.8 to +11.6; p = 0.409). Median number of dissected lymph node stations was 4 in both groups, and nodal upstaging was 8.1% versus 8.2%. At a median follow-up of 918 days, two-year survival was 91% and 90.3%. Case allocation was concentrated, with 53 of the 100 training cases performed by a single registrar; excluding these, major morbidity was 6.4% versus 7.8%, while the adjusted operative time difference was +8.9% (95% CI -0.9 to +19.7; p = 0.076). CONCLUSIONS: No precise estimate of the independent association between trainee participation and outcomes can be obtained from this selected cohort. Clinically important differences cannot be excluded given the cohort size and the selection of less complex cases for training. This programme was delivered with careful case selection and continuous consultant supervision throughout.